Skip to main content

​
Direct answer

Provider metrics can differ because the observed answers, model, source behavior, eligible denominator, or prompt coverage differs. Qwairy can show where the difference occurs, but it cannot reveal a provider’s private training data or ranking logic.

​
Data required

  • Matched answered prompts across the providers or models
  • Provider and model identifiers
  • Identical period, country, language, topic, tag, and funnel filters
  • Raw numerators and denominators for each metric
  • Representative answers and citations from each provider

​
Workflow

  1. In Cockpit > Overview, set a fixed scope and compare one metric at a time.
  2. Confirm that each provider has comparable completed-answer and answered-prompt counts.
  3. In Cockpit > GEO Matrix, locate the prompts and topics producing the gap.
  4. Separate models within each provider rather than merging them under a family label.
  5. Read the underlying answers in Monitor > Response Analysis and inspect sources in Monitor > Source Explorer.
  6. Recalculate the comparison after excluding missing or unmatched observations.

​
Interpretation

Product Mention Rate and Citation Rate are response-level and use different eligible answer denominators. Product Coverage is prompt-level. Product Share of Voice counts mention occurrences. A provider can therefore rank differently on each metric without any inconsistency. Prefer raw components over a blended score when diagnosing a difference. A small denominator can amplify one answer, and missing answers are not brand absences.

​
Possible next actions

  • Test whether the gap persists on a matched prompt cohort.
  • Investigate a topic or model that accounts for most of the difference.
  • Review source differences as a possible explanation without treating them as proven causes.
  • Adjust monitoring when one provider has incomplete coverage, then collect a comparable sample.
Document the hypothesis and preserve the scope for the next comparison.

​
Limitations

  • Provider output can vary between runs.
  • Provider and model availability can differ by market.
  • Qwairy observes outputs and exposed citations, not private provider systems.
  • Aggregate scores can hide metric-specific and prompt-specific differences.